[CI] Support using blob prefix for downloading

So that uploading and downloading use the same file paths. Also refactor
everything so that _get_blob_prefix is the common implementation.

.ci/cache_lit_timing_files.py

diff --git a/.ci/cache_lit_timing_files.py b/.ci/cache_lit_timing_files.py
index b44b826..ca223c2 100644
--- a/.ci/cache_lit_timing_files.py
+++ b/.ci/cache_lit_timing_files.py
@@ -22,10 +22,13 @@
 GCS_PARALLELISM = 100
 
 
+def _get_blob_prefix():
+    return f"lit_timing_{platform.system().lower()}"
+
+
 def _maybe_upload_timing_file(bucket, timing_file_path):
-    blob_prefix = f"lit_timing_{platform.system().lower()}/"
     if os.path.exists(timing_file_path):
-        timing_file_blob = bucket.blob(blob_prefix + timing_file_path)
+        timing_file_blob = bucket.blob(_get_blob_prefix() + "/" + timing_file_path)
         timing_file_blob.upload_from_filename(timing_file_path)
 
 
@@ -45,14 +48,14 @@
 
 
 def _maybe_download_timing_file(blob):
-    file_name = blob.name.removeprefix("lit_timing/")
+    file_name = blob.name.removeprefix(_get_blob_prefix() + "/")
     pathlib.Path(os.path.dirname(file_name)).mkdir(parents=True, exist_ok=True)
     blob.download_to_filename(file_name)
 
 
 def download_timing_files(storage_client, bucket_name: str):
     bucket = storage_client.bucket(bucket_name)
-    blobs = bucket.list_blobs(prefix="lit_timing")
+    blobs = bucket.list_blobs(prefix=_get_blob_prefix())
     with multiprocessing.pool.ThreadPool(GCS_PARALLELISM) as thread_pool:
         futures = []
         for timing_file_blob in blobs: